4

My other Question about "Add missing xts/zoo data with linear interpolation in R" you can find here Add missing xts/zoo data with linear interpolation in R .

But in general there is one more problem with my data - I do have "wrong" values not making sense:

"2012-04-09 05:03:00",2
"2012-04-09 05:04:00",4
"2012-04-09 05:05:39",-10
"2012-04-09 05:09:00",0
"2012-04-09 05:10:00",1

So the replacement of missing dates works:

y <- merge(y, zoo(,seq(start(y),end(y),by="min")), all=TRUE)
y <- na.approx(y)

But as you can see the -10 makes no sense, and the value is not at min:sec with value 00. I need a solution like na.rm. Thanks!

Community
  • 1
  • 1
Herr Student
  • 853
  • 14
  • 26

2 Answers2

4

It is not clear what do you want to do. But I guess you want to remove some outliers from xts object. If you want a solution like "na.rm", one idea is to replace non desired values by NA then you remove them using na.omit.

x <- read.zoo(text='
"2012-04-09 05:03:00",2
"2012-04-09 05:04:00",4
"2012-04-09 05:05:39",-10
"2012-04-09 05:09:00",0
"2012-04-09 05:10:00",1',sep=',',tz='')

x[x == -10] <- NA
na.omit(x)

                    x
2012-04-09 05:03:00 2
2012-04-09 05:04:00 4
2012-04-09 05:09:00 0
2012-04-09 05:10:00 1

EDIT

To get condition per date , you can look at index(x) and format it for example. 

format(index(dat),'%S')
[1] "00" "00" "39" "00" "00"

But here I use built-in .indexsec ( see also .indexmin, .indexhour,..)

dat[.indexsec(dat) != 0]
2012-04-09 05:05:39 
                -10
agstudy
  • 119,832
  • 17
  • 199
  • 261
  • Exactly what I mean - thank you! This solution is by looking for a specific value - is there also a possibility by date? Not having h:m:00 - for seconds? – Herr Student Apr 15 '13 at 13:13
4

Here's how to replace value that are below a threshold (0 in this case) with NA. You can run na.approx afterwards.

# read in
Lines <- '"2012-04-09 05:03:00",2
"2012-04-09 05:04:00",4
"2012-04-09 05:05:39",-10
"2012-04-09 05:09:00",0
"2012-04-09 05:10:00",1
'
data2 <- as.xts(read.zoo(text = Lines, sep = ",", tz = ""))

# perform calculation
data2[data2<0] <- NA
G. Grothendieck
  • 254,981
  • 17
  • 203
  • 341
Pierre Lapointe
  • 16,017
  • 2
  • 43
  • 56